Buying AI: How to Evaluate a Tool or a Vendor Before You Pay

Every week another "top ten AI tools" list appears. **And six months later half the list has

changed** — and you paid for one of them and do not use it.

This book is not another list. It is the method you vet any tool with, including ones not built yet.

The sentence this book rests on

**You are not buying a tool — you are buying a change in how you work. And the subscription is its cheapest

part.**

Read it twice.

Because when you subscribe to a tool, the meter does not start at the price. It starts at: who will learn

it, which process changes, who reviews its output, how long until it works, **and what happens the day you

want out**.

And the sticker price is the smallest number in the equation — we calculate the rest in chapter five.

When not to buy — honestly

I sell systems and build tools. And I will tell you when not to pay:

• When you cannot write the problem with a number. "Our work is slow" is not a problem, it is a feeling.
• When the process was never decided. The tool will freeze the chaos, not fix it.
• When you have not exhausted the free tier. Many tools give a small business enough for nothing.
• When the task is rare. Four times a month does not justify a subscription, a learning curve and

maintenance.

• When the motive is that everyone is buying.
• And when the demo impressed you before you knew your problem — the most common of all.

And the last deserves a name:

A solution looking for a problem.

You saw an impressive demo, and started hunting through your business for somewhere to put it. **The order is

inverted**: the problem first, then the search, then the demo — not the reverse.

The rule that condenses the book

Write the problem before you write the tool's name.

And if you cannot write it in one sentence with a number, you are not ready to buy. Chapter two is

entirely about that number.

Four questions before you open any pricing page

1. What is the problem, as a number?

"We spend six hours a week transcribing orders by hand" ← a problem.

"We want to benefit from AI" ← not a problem.

2. What do we do instead today?

Because what you do today is the baseline, and it is what you will compare against.

3. Why now?

Has something changed — volume, an employee leaving, a new customer? Or did you see an advertisement?

4. What will change in how we work?

Who will use it, which step disappears, and who reviews the output.

And the fourth reveals the real size of the decision — and most buyers only answer it after paying.

What this book is — and is not

It is about the purchase decision: how to measure your problem, vet a claim, calculate the three-year

cost, design a pilot, write a contract, measure afterwards, and exit if you must.

It is not about what may be put into the tool. That is a data and privacy question with its own course

("Your Data and AI") containing twelve questions for any tool. **I point at them in chapter six and will

not repeat them — and I add what belongs to purchasing alone: what you get in writing before signing.**

And it is not about automation. If you are building a process that runs without you opening it, that is

another book ("Automating Your Work with AI").

And I will not name a tool or a vendor. Names change and offers shift, **and the questions you vet a

vendor with do not** — and those are what you leave chapter four with.

Action steps

1. Write the sentence and post it: you are buying a change in work, not a tool.
2. Write your problem in one sentence with a number — and if you cannot, stop here.
3. Answer the four questions in writing before any pricing page.
4. Ask honestly: did I start with the problem or with the demo?
5. Go through the "when not to buy" list — and if an item applies, do not buy now.
6. Do not book a demo before chapter four.